sporely.
Get started free

Your Traffic Chart Can't Tell You What Worked

By Görkem Sarptunalı · Last updated · 5 min read

Key takeaways

  • Write each action down the day it starts, not when someone asks about the spike.
  • Give every action one stage it should move and one signal that shows its result.
  • Tie the signal to the action itself: a UTM campaign, a CRM source value, a form, a code.
  • Log the changes outside marketing too: pricing, sales hires, outages, site releases.
  • Treat always-on programmes differently from experiments: monthly targets for one, a deadline and a call for the other.
Your traffic chart can't tell you why A rising line over a time axis, with small mushrooms on the axis marking the actions taken: a price page change, a webinar and LinkedIn ads. The rise starts after the webinar, which only shows because the actions were logged. Your traffic chartcan't tell you why.Log the actions behind the line,then read the line again.Price pageWebinarLinkedIn ads View full size

Someone asks what you changed in March, because March was the best month of the year. The team remembers a webinar, a new ad set and maybe a pricing page tweak, but nobody is sure of the order, and the person who launched the ads has since left. Most teams have this conversation every quarter, and it always ends with the most confident memory winning.

Traffic is the effect, actions are the cause

Analytics tools are very good at recording what visitors did. They don't record what you did. The webinar, the new form, the partner placement and the email you stopped sending all happened outside the chart, so the chart can only show a line going up or down.

Without a record of actions, every movement gets explained after the fact, and after-the-fact stories are generous to whatever the storyteller worked on. With a record, you can ask a narrower question: which change started just before the line moved, on the stage it was meant to move?

What to write down for each action

Keep it to one line per action. Long forms don't get filled in.

  • What changed, in plain words: "Shorter demo form, 7 fields to 3".
  • The stage it should move: lead, qualified lead, demo, customer. One stage, the closest one to the action.
  • When it started, and when it ended if it has an end.
  • How you'll see its result: the signal that belongs to this action and nothing else.
  • What success looks like, decided before you see any numbers: "+20% demo request rate within six weeks".
  • Who owns it, so there's someone to ask in three months.
One line per action An action log with five illustrative rows: a partner banner, a webinar, a shorter demo form, always-on LinkedIn content and a pricing page rewrite. Each row lists the stage it should move, when it started, the signal that shows its result, its target and the current call, such as working, too early or not working. One line per actionWhat changed, which stage it should move, how you will see it, and the call. Illustrative rows.ActionStageStartedHow you'll see itTargetCallPartner banner on a fintech siteLeadSep 29sub-channel = partner name+3 a weekWorkingWebinar on Islamic bankingMQLOct 6form = webinar sign-up40 sign-upsToo earlyShorter demo formSQLOct 1demo requests / visits+20% rateNot workingLinkedIn content (always-on)LeadMonthlyutm_source = linkedin20 a monthOn pacePricing page rewriteCustomerOct 13trial to paid rate+5 pointsToo early View full size

Give each action its own signal

An action without its own signal can only be judged by the total, and totals move for many reasons at once. The trick is to connect the action to something only it produces.

  • Links you control: a dedicated utm_campaign value for every campaign, newsletter and partner placement.
  • CRM source values: many CRMs record where a contact came from with extra detail. HubSpot's Original Traffic Source has drill-down properties for things like the campaign name or the entrance URL (HubSpot's guide). A partner, event or sub-channel value is often all you need.
  • Forms and pages: a separate form or landing page for the webinar, so sign-ups count themselves.
  • Codes: a discount or referral code per partner or podcast.

When none of these is possible, for a brand campaign or a pricing change, say so in the log and judge it on the stage as a whole, with lower confidence.

Write down what isn't marketing

The biggest surprises in a funnel are often not campaigns. A price change, a new salesperson, a site migration, a two-day outage or a competitor's launch can move the same numbers your campaign is supposed to move. If they're not in the log, the campaign gets the credit or the blame. One line each is enough, with a date.

Programmes and experiments are different animals

Some work never ends: weekly LinkedIn posts, SEO, a monthly newsletter. Judging these like a test with a deadline produces "too early" forever. Give them a monthly target instead, and look at the trend month by month.

Other work is a bet: a new channel, a new offer, a redesigned form. These need a hypothesis, a fair test window and a call at the end, keep, change or stop. Mixing the two in one list is fine, as long as each line says which kind it is.

The weekly review, in fifteen minutes

  1. Add anything that started or stopped this week, including the non-marketing changes.
  2. For each live action, check its own signal against the target and against the stage's usual range. The campaign guide covers how to tell a real change from an ordinary week.
  3. Look one stage further. More leads that never become qualified leads is volume, not growth.
  4. Note the call: working, not working, too early, or not enough data yet.
  5. Flag actions that started within a week or two of each other on the same stage; their credit is shared until you can separate them.

If you use Google Analytics, its annotations are a useful mirror of the log: you can mark a date or a date range on any line graph, and the note shows on every report with one (Google's guide to annotations). They're good for "what happened on this date". They don't hold a target, a signal or a verdict, so keep the log itself somewhere the whole team can see.

Common traps

  • Logging after the fact. Entries written once the result is known tend to describe the result, not the intent.
  • Only logging the winners. A log of successes can't teach you anything about what doesn't work for your audience.
  • One action, five stages. If a change "should help everything", pick the stage closest to it and judge it there.
  • Logging every tweak. A new ad image isn't an action; a new audience or offer is. Keep the log readable.

Quick Q&A

What is a marketing action log?

A running list of every change the team makes, one line each, with the date it started, the funnel stage it should move, the signal that shows its result and what success looks like.

How is this different from GA4 annotations?

Annotations mark a date on a chart. An action log also records the stage, the signal, the target and the verdict, and covers changes outside your website.

Use the closest signal that only it produces, such as a CRM source value, a dedicated form or a code. If there's none, judge it on the whole stage and treat the result with lower confidence.

How often should we review it?

Weekly is enough for most teams. Each review adds new actions, checks live ones against their targets and makes a call on those that have had a fair run.

What if two actions start at the same time?

Note it in both lines and treat the credit as shared. Look for the one whose effect matches its own stage and timing, and record the result as unclear if they can't be separated.

Keep reading

More from the blog